Artificial bee colony (ABC) optimized edge potential function (EPF) approach to target recognition for low-altitude aircraft
Introduction
Target recognition is a key issue to achieve autonomous reconnaissance and attack for aircraft at low altitude. In many countries, target recognition technology for aircrafts at low altitude is highly confidential, and it is accordingly difficult to see its specific technical details (Shi et al., 2006). In order to obtain accurate identification results and hence to meet the practical requirements of aircrafts reconnaissance system, the proposed target recognition method must be efficient, stable, and convenient for promotion (Zhou et al., 2009, Francisco et al., 2009). Among all the methods, shape representation and matching is a very important aspect, and has been extensively used for solving object recognition problem (Scasellati and Alexopoulos, 1994, Belongie et al., 2002).
Generally, shape matching schemes involve two general steps: feature extraction, and similarity measuring (Veltkamp, 2001). Various methods have been used to determine the similarity between planar shapes, including moment-based matching (Hu, 1962, Taubin and Copper, 1991), Hausdorff distance based matching (Saber and Tekalp, 1997, Huttenlocker et al., 1993), and so on. Edge Potential Function (EPF) is a newly-developed similarity evaluating measure, which was firstly proposed by Minh-Son et al. (2007). This conception is derived from the potential generated by charged particles and has been proved its feasibility and reliability over Hausdorff distance and Chamfer distance measures.
Artificial bee colony (ABC) algorithm is a new optimization method, which is based on swarm intelligence and motivated by the intelligent behavior of honey bees. ABC algorithm has been proved to possess a better performance in function optimization problem, compared with genetic algorithm (GA), differential evolution (DE) and particle swarm optimization (PSO) (Karaboga and Basturk, 2007, Karaboga and Basturk, 2008). The main advantage of ABC algorithm lies in that it conducts local search in each iteration, thus the probability of finding the optimal results is significantly increased, which can efficiently avoid local optimum to a larger extent.
In this work, the EPF is adopted to provide a type of attractive pattern for a matching contour, which is conveniently exploited by ABC algorithm. In this way, the best match can be obtained when the sketch image translates, reorients and scales itself to maximize the potential value. The convergence for ABC algorithm is also proved theoretically.
The remainder of this paper is organized as follows. Section 2 introduces the principle of EPF. Section 3 describes the basic principle and implementation procedure of ABC algorithm in detail, and the convergence proof of the ABC algorithm is also presented in this section. Section 4 proposes our ABC optimized EPF approach to target recognition task. Then, in Section 5, series of comparison experiments are conducted to verify the feasibility and effectiveness of our proposed approach over the traditional genetic algorithm. Our concluding remarks and future work are contained in the final section.
Section snippets
The principle of EPF
EPF was firstly put forward by Minh-Son et al. (2007). This conception was derived from the potential generated by charged particles, and was especially adopted to model the attraction generated by edge structures contained in an image over similar curves.
A set of point charges in a homogeneous background can generate a potential, the intensity of which depends on the distance from the charges and the electrical permittivity of the medium , namelywhere and are
ABC algorithm
ABC algorithm was firstly proposed by simulating the self-organization simulation model of honey bees (Seeley, 1995). In this model, although each bee only performs one single task, yet through a variety of information communication ways between bees, the entire colony can complete a number of complex works such as hives building, pollen harvest and so on. Then in 2003, Dušan Teodorović further introduced a bee colony optimization (BCO) algorithm (Teodorovic and Orco, 2005). Then, Dervis
ABC optimized EPF approach to target identify
The implementation procedure of our proposed ABC optimized EPF approach to target recognition for aircraft at low altitude can be described as follows:
- Step 1
Image pre-processing
- (1)
Obtain the image, and convert it into grayscale format for further edge detection operation.
- (2)
Filter the target image to remove the noise.
Conduct filtering operation to the obtained grayscale image in order to mitigate the effect of noise. For this purpose, we applied the median filtering method, which was certified to have a
- (1)
Experimental results and analysis
In order to investigate the feasibility and effectiveness of the proposed method in this work, series of experiments are conducted, and further comparative experimental results with the GA method is also given.
The initial parameters of ABC algorithm were set as: , , , , .
The first experiment (Case 1) is to find an isosceles triangle among a variety of shapes in the original image. After a 330 degree rotation, 1.2 times scaling, and a [78, 68] translation, the
Concluding remarks
As the target recognition for aircraft at low altitude is rather complicated, a novel ABC optimized EPF approach to target identification for aircraft at low altitude is proposed in this paper. This hybrid method takes advantages of the accuracy and stability for EPF in target shape recognition, and ABC algorithm is adopted to optimize the matching parameters. Series of experiments are conducted, and experimental comparison results between the proposed method and the traditional GA are also
Acknowledgements
This work was partially supported by the National High Technology Research and Development Program of China (863 Program), Natural Science Foundation of China (NSFC) under grant #60975072 and #60604009, Aeronautical Science Foundation of China (Key Program) under grant #2008ZC01006, “Beijng NOVA Program” Foundation of China under grant #2007A0017, Program for New Century Excellent Talents in University (NCET) of China, and Graduate Innovation Practice Foundation of Beihang University, China.
The
References (19)
- et al.
Application of honey bee mating optimization algorithm on clustering
Appl. Math. Comput.
(2007) - et al.
On the performance of artificial bee colony (ABC) algorithm
Appl. Soft Comput.
(2008) - et al.
Region-based shape matching for automatic image annotation and query by example, J
Vis. Comm. Image Represent.
(1997) An artificial bee colony algorithm for the leaf-constrained minimum spanning tree problem
Appl. Soft Comput.
(2009)- et al.
Shape matching and object recognition using shape contexts
IEEE Trans. Patten Anal. Machine Intell.
(2002) - et al.
A hybrid learning approach to tissue recognition in wound images
Internat. J. Intell. Comput. Cybernet.
(2009) Visual pattern recognition by moment invariants
IRE Trans. Inform. Theory
(1962)- et al.
Comparing images using the Hausdorff distance
Trans. Patten Anal. Machine Intell.
(1993) The evolution of the organization of work in social insects
Monit Zool., Italy
(1986)
Cited by (125)
A benchmark data set for aircraft type recognition from remote sensing images
2020, Applied Soft Computing JournalCitation Excerpt :In [3], a coarse-to-fine method is built to integrate high-level information for aircraft type recognition. A method using the artificial bee colony algorithm with an edge potential function is proposed in [14] to solve this problem. These methods for object type identification from remote sensing images have achieved significant results, but there are still many shortcomings.
A novel firefly algorithm based on gender difference and its convergence
2019, Applied Soft Computing JournalARNet: Prior Knowledge Reasoning Network for Aircraft Detection in Remote-Sensing Images
2024, IEEE Transactions on Geoscience and Remote SensingRemote Sensing Object Detection Meets Deep Learning: A metareview of challenges and advances
2023, IEEE Geoscience and Remote Sensing MagazineA benchmark dataset for deep learning-based airplane detection: HRPlanes
2023, International Journal of Engineering and Geosciences